Suspicious Crime Identification and Detection Based on Social Media Crime Analysis Using Machine Learning Algorithms

Abstract

Social networking sites, such as Facebook and Twitter, make it possible for people to communicate with each other. However, some people have exploited the influence of social networking sites for illegal activities, such as intimidating news stories and suspicious posts on local community websites. The main objective of this work is to identify suspicious posting activities on Facebook and predict crime rates among those posts. By using pre-processing steps, the dataset can be cleaned up for identification by eliminating missing values, avoiding duplicates, stemming, removing stop words from posts that may indicate a suspicious post, and encoding the dataset for common format conversion. In order to construct a suspected profile, posts from social media are clustered together. In the next step, relevant features related to crime are selected for prediction using a feature selection strategy. To predict the single crime that relates to the various suspicious activities, classification approaches will be applied.